Overview & Experiments 61 Synthesis Roadmap Lookahead Audit

Hold-out · adx_dmi_regime on unseen windows

Inconclusive

2026-06-14 adx_dmi_regimeholdoutvalidationpromotion

Hypothesis

The fixed finalist params (adx_dmi_regime+fixed_stop, 4h) from sweep MS_20260606_043928 remain profitable out-of-selection: return_pct > 0 in >= 3/4 years the sweep never saw.

Verdict

**BORDERLINE** — only 2/4 unseen windows profitable (3 required). Partly selection; only with caution / small size.

cfg

{'name': 'ADX_DMI_REGIME_4H', 'signal': 'adx_dmi_regime', 'interval': '4h', 'conditions': ['adx_dmi_regime', 'fixed_stop'], 'fixed_stop_pct': 1.5, 'regime_range_enter': 22.0, 'regime_trend_enter': 30.0}

gate

return_pct > 0 in >= 3/4 unseen windows

sweep

MS_20260606_043928 / P2_00124

windows

['Y2020', 'Y2023', 'Y2024', 'Y2025']

holdout_wins

2/4

Hold-out · adx_dmi_regime on unseen windows

2026-06-14 · status: inconclusive · 1.8s

Hypothesis: The fixed finalist params (adx_dmi_regime+fixed_stop, 4h) from sweep MS_20260606_043928 remain profitable out-of-selection: return_pct > 0 in >= 3/4 years the sweep never saw.

Verdict: BORDERLINE — only 2/4 unseen windows profitable (3 required). Partly selection; only with caution / small size.

Key metrics

metric	value
holdout_wins	`2/4`
cfg	`{'name': 'ADX_DMI_REGIME_4H', 'signal': 'adx_dmi_regime', 'conditions': ['adx_dmi_regime', 'fixed_stop'], 'interval': '4h', 'regime_trend_enter': 30.0, 'regime_range_enter': 22.0, 'fixed_stop_pct': 1.5}`
windows	`['Y2020', 'Y2023', 'Y2024', 'Y2025']`
sweep	`MS_20260606_043928 / P2_00124`
gate	`return_pct > 0 in >= 3/4 unseen windows`

Hold-out · unseen years (never used by the sweep)

window	trades	return_pct	max_dd_pct	profit_factor	win
Y2020	56	-4.38	23.03	0.93	LOSS
Y2023	30	109.42	9.53	3.07	WIN
Y2024	51	35.34	21.11	1.65	WIN
Y2025	38	-12.28	16.69	0.69	LOSS